Download Hourly Energy Data from Powerwall

[atj777]

Is it possible to download the hourly energy (kWh) data from the Powerwall?

If I use the app and look at the data for a day, there's an option to download the data, however, the data are power (kW) at a point in time (5 minute intervals) and there's no easy way to to see how much energy (kWh) is being produced or consumed per hour.

If I look at the weekly data I can download energy, but it is whole day and not by hour.

 

[paulp]

Is it possible to download the hourly energy (kWh) data from the Powerwall?

If I use the app and look at the data for a day, there's an option to download the data, however, the data are power (kW) at a point in time (5 minute intervals) and there's no easy way to to see how much energy (kWh) is being produced or consumed per hour.

If I look at the weekly data I can download energy, but it is whole day and not by hour.

It downloads in xcel so you’d need to do a new auto calculate column converting the kw to kwh for each 5 minute interval, and then total them up at the end of the hour. Once you’ve setup one hour you should be able to copy and paste your formula for any other hour

 

[atj777]

It downloads in xcel so you’d need to do a new auto calculate column converting the kw to kwh for each 5 minute interval, and then total them up at the end of the hour. Once you’ve setup one hour you should be able to copy and paste your formula for any other hour

Are you sure you can do this?

It is my understanding it is displaying what power production/consumption is at that point in time, not the average over the 5 minute period.

My AC can cut in for just 3 minutes, or I can run the microwave for one minute. If either (or both) happen to be running on the 5 minute it will get counted. If they weren't, they won't get counted.

If I sum all the data for the day and then divide by 12 It comes sort of close to the numbers shown in the app, but not close enough to give me confidence. If data were truly average for the 5 minute period the calculations should be near exact.

For example, Home usage of 41.4 kWh versus 40.5 kWh.

 

[paulp]

Are you sure you can do this?

It is my understanding it is displaying what power production/consumption is at that point in time, not the average over the 5 minute period.

My AC can cut in for just 3 minutes, or I can run the microwave for one minute. If either (or both) happen to be running on the 5 minute it will get counted. If they weren't, they won't get counted.

If I sum all the data for the day and then divide by 12 It comes sort of close to the numbers shown in the app, but not close enough to give me confidence. If data were truly average for the 5 minute period the calculations should be near exact.

For example, Home usage of 41.4 kWh versus 40.5 kWh.

Good point. I’m not sure. Tesla have a lot more data than we see, which I found out during my recent interactions with them

 

[Vostok]

It is my understanding it is displaying what power production/consumption is at that point in time, not the average over the 5 minute period.

My AC can cut in for just 3 minutes, or I can run the microwave for one minute. If either (or both) happen to be running on the 5 minute it will get counted. If they weren't, they won't get counted.

If I sum all the data for the day and then divide by 12 It comes sort of close to the numbers shown in the app, but not close enough to give me confidence. If data were truly average for the 5 minute period the calculations should be near exact.

I don’t know for sure, but I suspect each data point is the instantaneous kW at the time stated in each row, and not an average for the 5 minutes leading up to that. So there will be swings and roundabouts, loads that only run for a few minutes will sometimes be active at the time the sample is taken, and not active at others. Given sufficient randomness in when that occurs, it should even out.

I think the bigger problem is that the data is only provided to one decimal place, and I don’t know how they do the rounding. That probably leads to larger systematic bias when you start to add up hundreds of numbers. I did suggest to Tesla last year they should provide it to 2dp, which would reduce any bias, but clearly that hasn’t happened yet.

In response to your original question, it depends on how fluent you are in Excel / coding. You could calculate hourly data with cell formulas, or if this is something you are keen to do on an ongoing basis, and can code, write Visual Basic procedures (macros) to do the calculations.

I’m a data nerd and have downloaded the data every day for nearly 2 years, I now have nearly a quarter of a million rows of data that I occasionally crunch in various ways.

 

[paulp]

I don’t know for sure, but I suspect each data point is the instantaneous kW at the time stated in each row, and not an average for the 5 minutes leading up to that. So there will be swings and roundabouts, loads that only run for a few minutes will sometimes be active at the time the sample is taken, and not active at others. Given sufficient randomness in when that occurs, it should even out.

I think the bigger problem is that the data is only provided to one decimal place, and I don’t know how they do the rounding. That probably leads to larger systematic bias when you start to add up hundreds of numbers. I did suggest to Tesla last year they should provide it to 2dp, which would reduce any bias, but clearly that hasn’t happened yet.

In response to your original question, it depends on how fluent you are in Excel / coding. You could calculate hourly data with cell formulas, or if this is something you are keen to do on an ongoing basis, and can code, write Visual Basic procedures (macros) to do the calculations.

I’m a data nerd and have downloaded the data every day for nearly 2 years, I now have nearly a quarter of a million rows of data that I occasionally crunch in various ways.

I thought I was OTT downloading it monthly. Daily is a whole new level.

 

[Vostok]

I thought I was OTT downloading it monthly. Daily is a whole new level.

Yeah, it is, although I don't actually do it every day, every couple of weeks when I otherwise have an idle moment I scroll back and download each day individually and add the CSVs to my Notes App. When I have the whole month, I import them into Excel.

I also suggested to Tesla that they have an additional "download detailed data" button when showing monthly or yearly views, so that you could download the 5-minute data for that entire period in one go, instead of just the summary data for that view. That would make life so much easier. They haven't done that yet either.

I logged into the PW2 once to see if I could write a script to pull this data automatically (and with more resolution), but the unoffocial API document I found did not appear to offer that capability. It also warned you that if your screwed up the system by pushing commands to the PW2 that you shouldn't, bad things could happen. So I retreated.

 

[atj777]

In response to your original question, it depends on how fluent you are in Excel / coding. You could calculate hourly data with cell formulas, or if this is something you are keen to do on an ongoing basis, and can code, write Visual Basic procedures (macros) to do the calculations.

It is more so I can verify (or at least come close to verifying) my bills from Powershop.

In theory, I switched to ToU billing from 1 June from a fixed rate. The Powershop app shows me my daily usage and estimate of the costs. At the moment the costs it is showing me still reflects the fixed rate but I want to be ready for when they fix that (or be able to reasonably argue with them if they don't fix it).

It is easy enough to me to use Numbers (rather than Excel as I've downloaded the data to iCloud) and just add two extra columns: one to add up the grid kW for each time period and divide by 12; one column to multiple by the rate for that period.

As you say, if it is point of time, it will probably be swings and roundabouts and will give me a close enough estimate.

What is interesting is that I just realised that the Photoshop app lets me get the usage for a specific time period (and it seems reasonably close to the numbers downloaded from my gateway).

 

[paulp]

It is more so I can verify (or at least come close to verifying) my bills from Powershop.

In theory, I switched to ToU billing from 1 June from a fixed rate. The Powershop app shows me my daily usage and estimate of the costs. At the moment the costs it is showing me still reflects the fixed rate but I want to be ready for when they fix that (or be able to reasonably argue with them if they don't fix it).

It is easy enough to me to use Numbers (rather than Excel as I've downloaded the data to iCloud) and just add two extra columns: one to add up the grid kW for each time period and divide by 12; one column to multiple by the rate for that period.

As you say, if it is point of time, it will probably be swings and roundabouts and will give me a close enough estimate.

What is interesting is that I just realised that the Photoshop app lets me get the usage for a specific time period (and it seems reasonably close to the numbers downloaded from my gateway).

I think you will need to convert the kw to kwh before you add them

 

[atj777]

I think you will need to convert the kw to kwh before you add them

I can do it in either order. "convert the kW to kWh" is divide by 12 (5 minutes * 12 = 60 minutes)

(A+B+C+D+E)/12 = A/12 + B/12 + C/12 + D/12 + E/12.

 

[alanjackson]

I've tried to download but the button doesn't seem to do anything. What should I see? And where does it put the data?

 

[jamesgo]

Depending on your power company, if you're on a virtual power plant, eg AGL VPP, you can normally download the data easily from them (they query the powerwall every few seconds).

 

[WoodWombat]

I've tried to download but the button doesn't seem to do anything. What should I see? And where does it put the data?

I suspect that the app function is currently broken on the Android app since it works OK on my iPhone.

As OP stated, data is in kW, one can only hope that it is average kW over that time period, not instantaneous kW otherwise it is completely useless.

 

[paulp]

I suspect that the app function is currently broken on the Android app since it works OK on my iPhone.

As OP stated, data is in kW, one can only hope that it is average kW over that time period, not instantaneous kW otherwise it is completely useless.

It works on my iphone also

 

[Vostok]

As OP stated, data is in kW, one can only hope that it is average kW over that time period, not instantaneous kW otherwise it is completely useless.

I don’t know for sure, but I suspect it is the instantaneous value at the end of each 5-minute period. If you directly interrogate the Tesla API, the data is returned as a floating point number in Watts to an absurd number of decimal places (way beyond the point where it would be accurate). The Tesla App very annoyingly changes the Watts to kW and rounds the number to 1 decimal place if you download it that way.

If it is an instantaneous value, it doesn’t make it “completely useless”. Half the time that data point will be more than the 5-minute average, and the other half it will be less. Swings and roundabouts I reckon, with a mean error close to zero.

 

[JPFlathead]

I've done a download over the course of 6 months, using the daily extract at 5 minute intervals. the numbers balance well. That is, the power from the panels to the powerwall, used from the powerwall, and exchanged with the power company all balance to within <0.1% (i.e. rounding errors) and agree with the bills from the meter and power company. So it's most likely not instantaneous, but average over the 5 minute interval and gets close enough for estimating power usage.

I just wish there was a less manual method of getting the data out.

 

[enrgyscientist]

I've done a download over the course of 6 months, using the daily extract at 5 minute intervals. the numbers balance well. That is, the power from the panels to the powerwall, used from the powerwall, and exchanged with the power company all balance to within <0.1% (i.e. rounding errors) and agree with the bills from the meter and power company. So it's most likely not instantaneous, but average over the 5 minute interval and gets close enough for estimating power usage.

I just wish there was a less manual method of getting the data out.

Hi @JPFlathead ! Would you be willing to share your data with me? I am building whole home energy optimization models and need at least one month worth of granular Powerwall interval data for a simulation. Unfortunately I dont own a Powerwall myself. Your help would be greatly appreciated!!

 

[Vostok]

I just wish there was a less manual method of getting the data out.

There is a less manual way - use a Python script that talks to the Tesla Owner API and pulls the data down. There are a number of different ones on GitHub, TeslaPy is probably the most widely known one. However you do need to be a little bit coding literate to use them.

The other advantage of this method is you get the data to 3 decimal places of kilowatts instead of one decimal place via the manual download method.

I have all the 5 minute data since my PW2 was installed in 2019. Nearly half a million rows of data now. I also pulled out the historical State of Charge (SoC) data once Tesla introduced the feature in the App - and you could also pull that data right back to inception. So it must have always been there, the App just made it available as part of the update that introduced the daily PW2 SoC charts.

I have found my PW2 data has very close agreement with the solar export stated on my electricity bill, OK agreement with the electricity used at peak and off-peak times (within a few percent) but it has been way off with my shoulder usage by more than 10%. And I have no idea why that is.

 

[cafz]

but it has been way off with my shoulder usage by more than 10%. And I have no idea why that is.

Daylight saving time issue? Some retailers use AEST, some AEDT. And at least the part of the Tesla app that counts up how much you've spent charging the car at home doesn't cope well with the time change.

 

[Vostok]

Daylight saving time issue? Some retailers use AEST, some AEDT.

I don’t think so, because that would affect the comparison across all the tarriff periods, not just shoulder.